What is wrong with the quantitative standards for market risk?

Abstract

The purpose of this paper is to evaluate the quantitative standards laid down under the second Basel Accords for the implementation of internal market risk models by banks. The paper surveys available research to evaluate the standards. The standards don’t prescribe a VaR method despite evidence that volatility of financial returns is conditional and financial returns are fat tailed. The requirement of a minimum historical period also runs contrary to the finding that volatility is time varying and clustered resulting in banks being able to use weighting schemes conservatively only. The minimum horizon of ten days requires use of a scaling rule that is not accurate. The 99% confidence level requirement increases the inaccuracy when using a normal assumption on fat tailed data. The minimum updation period and minimum historical period requirements effectively smooth the market risk charge over and above the smoothing by the requirement of averaging VaR resulting in unresponsive market risk charges. The regulatory back testing framework is based on unconditional coverage and doesnot penalize clustered VaR exceptions.

Key Words: Basel accord, GARCH, Historical simulation, Market risk, Value-at-risk, Volatility, Conditional volatility, Back testing.